To date, PDAC remains a lethal malignancy in the digestive system. Though marked improvement was made in the treatment of other cancer types in recent years, including gastric cancer, colon-rectal cancer, and hepatic cancer, the successful therapeutic modality for these tumors cannot translate to improved survival in PDAC, one of the key links behind this is the heterogeneity in PDAC TME and TIME. Thus, progress in stratifying immune-relevant risk groups of PDAC is urgently needed. In this study, we established a scoring method (T-C score) based on the consensus clustering of the survival-related TLSs- and CD8 + T-cell-associated gene expression profiles in PDAC patients. The results of this study indicated that the T-C score could independently and reliably work as a prognostic factor, it can also serve as a predictive indicator to estimate the therapeutic responses to ICBs, small molecular drugs, and different chemotherapy regimens. The PDAC patients with low T-C scores were a more active immune phenotype with better survival and were marked with higher immune, stromal, and ESTIMATE scores, whereas higher genomic alterations were found in patients with high T-C scores. Furthermore, the patients in the low T-C score group were more responsive to the ICBs immunotherapy and chemotherapeutic agents like paclitaxel or 5-fluorouracil.
The recent decade has emphasized the relationship between tumor microenvironments and prognosis in solid cancers. Against such a background, researchers have focused mainly on the role of T cells, highlighting the association of tumor-infiltrating T lymphocytes with patient’s prognosis in numerous tumors, including PDAC [40–46]. However, recently, B-cells as well as TLSs were discovered as critical components in the TIME in terms of patient prognosis and response to immunotherapy [11–13, 44, 47]. TLSs represented a special TIME status and serve as a beacon for antigen presentation and T-cell activation. Previous publications demonstrated mature TLSs (M-TLSs) had sophisticated architecture which contains multiple immune cell types and structures including adjacent T- and B-cell zones, activated dendritic cells present within the T cell zone, germinal center (GC) reactions in the B cell follicle [11, 14, 48, 49], endothelial cells, and plasma cells. These were quite different from early-TLSs (E-TLSs) which predominantly consisted of T- and B-cells without sophisticated architecture. Since not all TLS manifested with the same histological architecture and phenotypic features, researchers have proposed a timeline for E-TLSs to M-TLSs transition [14, 48], which proposed different stages for TLSs in the midst of its maturation process [50]. Previous publications also demonstrated that different types of TLSs may exist in individual tumors independent of the spatial location of the TLS [12, 20], which indicated diversity in the immunological nature of TLSs in the tumor TIME, so there were researchers suggested the immunological phenotypes stimulated by the presence of TLSs may not be uniform, but instead be a deposition of anergic, terminally differentiated, exhausted or antigen-independent T-cell activation.
CD8 + T-cells were the most studied tumor-infiltrating immune cell type, for their important role in immune escape and immune treatment. according to their differentiation status, CD8 + T-cells can be subdivided into primitive T-cell, effector T-cell, and memory T-cell subsets. In the TME, the activation and suppression of these cells were controlled by various other cells and small molecules [6]. In addition to the various differentiations of CD8 + T-cells, the infiltration status of CD8 + T-cells could vary between tumors, even between patients who were plagued by the same tumor. The close relationship between CD8 + T-cells and TLSs were found in many cancers. Rita Cabrita et al [12]. found that the presence of TLSs was coupled with increased tumor-associated CD8 + T cells infiltration in melanoma, and the combination of infiltration of both TLSs and CD8 + T-cells was associated with better survival outcome compared to CD8 + T-cells infiltration alone. A positive correlation between CD8 + T-cell infiltration and TLSs presence was also observed in PDAC by Andrew J. Gunderson et al [9], they also found that the E-TLSs were correlated with greater CD8 + T-cell infiltration, whereas M-TLSs tumors were more enriched for activated CD4 + memory T-cells, which implied varied abilities in stimulating CD8 + T-cell infiltration at different stages of TLSs maturation.
Although the roles of TLSs and CD8 + T-cells in tumor biology and antitumor immunity have been widely studied, the potential clinical translation is still hampered by the heterogeneity of TLSs and their complex relationship with CD8 + T-cells. Hence, we focused mainly on developing a scoring method that utilized the gene signatures of both TLSs and CD8 + T-cells in PDAC to better stratify pancreatic cancer. In this study, the survival associated TLSs and CD8 + T-cells related genes were selected by the univariate Cox analysis, and based on that, PDAC samples were initially stratified into 2 T-C clusters based on consensus clustering. The patients in the T-C cluster-1 had relatively better survival and lower TMB compared to the T-C cluster-2 patients, which was consistent with the notion that mutation accumulation in cancer has a negative correlation with the OS of patients [35]. The T-C cluster-1 also had higher TLSs and CD8 + T-cell signature genes expression and higher immune cell infiltration than T-C cluster-2, These findings together suggested the TLSs and CD8 + T-cells might act synergistically to promote immune infiltration and patient survival, Therefore, an integrative characterization of the TLSs and CD8 + T-cells at gene level would be a better approach to stratify PDAC patients for further individualized assessment and therapy.
In the current study, the T-C score was established by consensus clustering and Boruta algorithm. Patients with lower T-C scores in TCGA-PAAD showed better survival, suggesting that the T-C score could work as a prognostic indicator. The predictive ability of the T-C score in PDAC for 1-, 2-, 3-, 4-, and 5-year survival was confirmed by the ROC curves. Further, the T-C score was verified as an independent risk factor for OS through the univariate and multivariate Cox regression analyses, the good predictive ability of the T-C score was further validated in the ICGC- CA, E-MTAB-6134, and the joint cohort of GSE71729 and GSE85916 via K-M analyses, time-dependent AUCs, univariate, and multivariate Cox regression analyses. Collectively, our current study confirmed the good predictive ability of the T-C score, which suggested the T-C score could be used in clinical practice in predicting OS for PDAC patients. PDAC is a genetic disease, The most prevalent form of genetic alteration in PDAC is single-nucleotide variants (SNVs) [51, 52]. KRAS, TP53, CDKN2A, and SMAD4 are among the earliest genetic variants in PDAC [53]. In the current research, though not all significant, higher mutation frequencies in KRAS (88% vs 31%), TP53 (75% vs 41%), CDKN2A (24% vs 12%), and SMAD4 (26% vs 19%) were found in high T-C score group, the mutation analysis revealed a positive correlation between the TMB and T-C score, as well as between CNV and T-C score. Notably, most of the patients in the T-C cluster-2 were in the high T-C score group. These findings suggested that gene mutations in PDAC might drive the profile of TLSs and CD8 + T-cell infiltration toward a pattern of high T-C score in PDAC. Stem-like cell phenotypes are closely related to many key steps in PDAC progression and the cancer metastatic cascade, as cancer cells expressing those features exhibit increased cell motility and invasiveness [54–56], therefore we estimated PDAC stemness using OCLR algorithm, and higher levels of stemness score were found in the high T-C score group, and a positive correlation was found between T-C score and stemness score, indicating more mutations in genes that encode oncogenes and epigenetic modifiers, frequent perturbations in specific mRNA/miRNA transcriptional networks, and more deregulation of signaling pathways for PDAC patients in the high T-C score group [31].
The tumor immune microenvironment (TIME) is closely related to the clinical prognosis of patients with tumors [57]. Different types of immune cell populations, such as lymphocytes and myeloid originated cells, comprised the major part of the TIME [58]. Through ssGSEA and CIBERSORTx, we recognized that the low T-C score group was a more immune-activated phenotype with significantly higher infiltrations of naïve B cells, CD8 + T cells, activated NK cells and monocytes, while the high T-C score group was heavily infiltrated by M0 macrophages, which was found to be an independent risk factor for various cancers [59–62]. Recent advances have indicated that the active pre-treatment immune status could have a positive impact on immunotherapy [63, 64], more Tumor-infiltrating CD8 + T cells in low T-C score group had marked influence on immunotherapy, as the classical mechanism behind the anti-PD-1/PD-L1 treatment is that the mutual suppression effect conducted by tumor cells and tumor-infiltrating CD8 + cytotoxic T cells (CTLs) could be abrogated by ICBs so that CD8 + CTLs could regain their cytotoxicity against tumor cells [65]. In light of the previously reported capability of TIME status in predicting sensitivity to ICBs treatment in patients with cancer [66], we analyzed the ICP genes expression, and a higher level of these genes expression was found in the low T-C score group, in accordance with this result, a significantly higher response rate to ICBs was observed in the same group in TIDE analysis. Though classically, the high TMB has been regarded as a predictive factor for the response to ICB treatment [67, 68], a different result was recently published regarding all solid cancer types [69]. Thus, the T-C score may serve as an alternative marker for predicting therapeutic response to ICBs. Emerging evidence has exhibited that multi-kinase inhibitors alone or synergistically used with HDAC inhibitors could suppress PDAC tumor growth [24, 70]. In the current study, the IC50 and CMap analysis demonstrated a possible higher sensitivity for the use of multi-kinase inhibitors and HDAC inhibitors in the low-risk group, which provided another therapeutic possibility for these PDAC patients.
There are some limitations in this study. First, batch effect may still occur even after correction due to different RNA sequencing methods and platforms. Second, as limited by lacking therapeutic data of a PDAC-related immunotherapy cohort, we used the TIDE web tool in speculating ICBs therapeutic response. Third, fields of clinical data in the ICGC-CA, E-MTAB-6134, and TCGA-PAAD were not exactly matching, which may affect verification effectiveness in multivariate cox analysis. Lastly, Large-scale prospective studies are required to further verify the value of the T-C score in predicting clinical benefits for PDAC patients.
In summary, We integrated transcriptome data and the expression of TLSs- and CD8 + T-cell-related genes to establish a new independent prognostic marker for PDAC, the T-C score, which could be used for the prediction of OS, immune infiltration status, tumor stemness, tumor mutation, and sensitivity to immune therapy and chemotherapeutics for PDAC.